A Lattice-based Method for Mining Frequent Weighted Itemsets Using Diffset

Mining Frequent Weighted Itemsets (FWIs) is an important problem in the field of data mining. Itemsets have different importance depending on their values in practical applications. Aiming at the problems of low efficiency in traditional pattern mining, a new method of weighted Node lattice (WN-lattice) in mining FWIs is proposed. This method includes three phases:(1) Building WN-lattice structure solve the high data compression. (2) A Diffset-based strategy for the efficient computation of the supports of inter-transaction patterns is described. Some theorems for pruning redundant itemsets quickly based on these theorems. (3) Frequent Weighted Itemsets Lattice Algorithm (FWIL) is described for directly mining FWIs using WN-lattice. The experimental results show the proposed approach FWIL is more efficient and more those used previously in terms of runtime, especially when run on very large and dense weighed databases.